Novel Process of Intrathymic Tumor-Immune Tolerance through CCR2-Mediated Recruitment of Sirpα+ Dendritic Cells: A Murine Model

Immune surveillance system can detect more efficiently secretory tumor-specific antigens, which are superior as a target for cancer immunotherapy. On the contrary, immune tolerance can be induced in the thymus when a tumor antigen is massively secreted into circulation. Thus, the secretion of tumor-specific antigen may have contradictory roles in tumor immunity in a context-dependent manner. However, it remains elusive on the precise cellular mechanism of intrathymic immune tolerance against tumor antigens. We previously demonstrated that a minor thymic conventional dendritic cell (cDC) subset, CD8α−Sirpα+ cDCs, but not the major subset, CD8α+Sirpα− cDCs can selectively capture blood-borne antigens and crucially contribute to the self-tolerance. In the present study, we further demonstrated that Sirpα+ cDCs can capture a blood-borne antigen leaking inside the interlobular vascular-rich regions (IVRs). Blood-borne antigen selectively captured by Sirpα+ cDCs can induce antigen-specific Treg generation or negative selection, depending on the immunogenicity of the presented antigen. Furthermore, CCR2 expression by thymic Sirpα+ cDCs and abundant expression of its ligands, particularly, CCL2 by tumor-bearing mice prompted us to examine the function of thymic Sirpα+ cDCs in tumor-bearing mice. Interestingly, tumor-bearing mice deposited CCL2 inside IVRs in the thymus. Moreover, tumor formation induced the accumulation of Sirpα+ cDCs in IVRs under the control of CCR2-CCL2 axis and enhanced their capacity to take up antigens, resulting in the shift from Treg differentiation to negative selection. Finally, intrathymic negative selection similarly ensued in CCR2-competent mice once the tumor-specific antigen was secreted into bloodstream. Thus, we demonstrated that thymic Sirpα+ cDCs crucially contribute to this novel process of intrathymic tumor immune tolerance.


Introduction
Immune surveillance system can detect more efficiently secretory tumor-specific antigens than intracellular ones. Thus, it is widely believed that secretory antigens are superior as a target for cancer immunotherapy to intracellular antigens [1,2]. On the contrary, immune tolerance can be induced when a tumor antigen is massively secreted into circulation. Supporting the latter notion, tumor antigen-specific CD4 + T cells were eliminated in the thymus and draining lymph nodes in the mice bearing a tumor, which constitutively secreted an antigen into the circulation [3,4]. Thus, a secretory tumor antigen may have contradictory roles in tumor immunity in a context-dependent manner. However, it remains elusive on the precise cellular mechanism and molecular regulation of immune tolerance against secretory tumor antigens.
Intrathymic educational processes consist of the induction of negative selection and natural Treg differentiation, and participate in orchestrating central immune tolerance, thereby reducing the risk of developing autoimmune disorders. Negative selection, a cardinal recessive tolerance process, is crucial for the maintenance of immune homeostasis while Treg-mediated immune regulation is a main gateway to dominant tolerance and can intervene in ongoing abnormal immune responses. Central tolerance is accurately executed by several antigen presenting cells (APCs) including medullary thymic epithelial cells (mTECs) and thymic DCs. A transcription factor, autoimmune regulator (AIRE) [5], is selectively expressed in mTECs and to a lower extent, thymic DCs, and constitutively regulates the transcription of tissuerestricted antigens in mTECs, thereby inducing central tolerance, negative selection and Treg differentiation from thymocytes which can recognize the antigens expressed in mTECs [5,6,7,8,9]. Additionally, some DCs can convey the peripheral self-antigens to the thymus, resulting in the induction of negative selection and Treg differentiation in the DC-dependent manner [10,11].
Recently, Wu and Shortman classified thymic DCs into three subsets, CD11c + CD11b 2 CD8a + Sirpa 2 and CD11c + CD11b + CD8a 2 Sirpa + cDCs, and CD11c + B220 + plasmacytoid DCs [12]. Among these three DC subsets, the most abundant CD8a + Sirpa 2 cDCs subset is clustered in the medulla with a low, but effective AIRE expression, and can present endogenous self-antigens. Moreover, they can take up self-antigens from mTECs and cross-present them to the thymocytes [12,13]. However, it still remains elusive on the intrathymic location and function of the other minor cDC, CD11c + CD11b + CD8a 2 Sirpa + subset. We previously demonstrated that most Sirpa + cDCs, as opposed to Sirpa 2 cDCs, are disseminated in close proximity to small vessels in the cortex and inside the perivascular regions [14]. Their unique intrathymic localization allows thymic Sirpa + cDCs to selectively capture intravenously injected antigens across the blood-thymus barrier [14]. Interestingly, mice deficient in a CC chemokine receptor, CCR2, exhibit a selective decrease in the Sirpa + cDC subset in the thymus, with an accumulation of T cells displaying reactivity against serum self-antigen in the periphery [14], suggesting the crucial contribution to the central tolerance against blood-borne antigens.
In the present study, more detailed histological analysis revealed that thymic Sirpa + cDCs were disseminated in the cortex and inside the interlobular vascular-rich regions (IVRs), encompassed with the cortical parenchyma. Moreover, Sirpa + cDCs efficiently captured blood-borne antigen leaking inside the IVRs in a CCR2dependent manner. Intravenously injected untreated ovalbumin (OVA) protein and highly immunogenic heat-aggregated OVA protein activated thymic Sirpa + cDCs to initiate antigen-specific Treg differentiation and negative selection for OVA-specific TCR transgenic thymocytes, respectively. Moreover, when developed subcutaneously, tumor induced the deposition of a CCR2 ligand, CCL2 inside IVRs, resulting in the enhancement of antigen uptake by intrathymic Sirpa + cDCs. We finally proved that the negative selection rather than Treg differentiation ensued in a CCR2-dependent manner once the tumor-specific antigen was secreted into bloodstream.

Tissues
Human tissues were obtained from autopsy with a written informed consent in accordance with the Declaration of Helsinki. All human studies were approved by the Medical Ethics Committee of Hokkaido University Graduate School of Medicine. Thymic tissues were flash-frozen in liquid nitrogen and stored at 280uC.
Plasmids pMYs-IRES-GFP and pLEGFP-N1 retrovirus vectors were purchased from Cell Biolabs and Clontech, respectively. An antagonistic form of CCL2 (7ND) is the amino terminus deleted form of human CCL2 and can exhibit antagonistic activities against CCR2 [15]. 7ND-expressing vector was prepared by cloning to the pMYs-IRES-GFP vector, the human CCL2 NH 2terminus-deleted cDNA linked with an epitope FLAG tag in the carboxyl terminal portion [16]. OVA cDNA with a truncated stop codon was amplified by PCR from chicken OVA cDNA and the signal peptide sequence of mouse albumin was synthesized as two complementary oligonucleotides (Gene Design). OVA cDNA linked with and without albumin signal peptide sequence were inserted into pLEGFP-N1 vector, to generate sOVA-GFP and OVA-GFP, respectively.

Transfectants
The Phoenix packaging cell line was transfected with 7ND, OVA-GFP, or sOVA-GFP retrovirus vector to produce the retrovirus in the culture supernatant, which was subjected to the infection. A murine colorectal cancer cell line, Col26 [17] was infected with the resultant retrovirus to generate clones expressing stably 7ND, OVA-GFP, or sOVA-GFP. The concentration of OVA protein in the culture supernatant (1610 5 cells/ml) was determined using Chicken Egg Ovalbumin ELISA kit (Alpha Diagnostic).

Mice
Specific pathogen-free male BALB/c and C57BL/6 mice were purchased from Charles River Japan, and designated as WT mice. CCR2 2/2 mice [18] were kindly provided by Dr. William Kuziel (University of Texas San Antonio), and were backcrossed to BALB/c mice for more than 8 generations. CCR2 2/2 backcrossed to C57BL/6 mice were provided also by Dr. Kuziel via Dr. Motohiro Takeya (Kumamoto University). DO11.10 mice expressing a transgenic TCR that recognizes the OVA 323-339 peptide in the context of I-A d were kindly provided by Dr. Yasunari Nakamoto (Kanazawa University) and were maintained as heterozygotes. OT-I and OT-II mice express transgenic TCRs that recognize the OVA 257-264 peptide in the context of H-2K b and OVA 323-339 peptide in the context of I-A b , respectively. These mice were supplied from CARD, Kumamoto University. DO11.10 mice were mated with CCR2 2/2 mice to generate DO11.10/CCR2 2/2 mice on a BALB/c background. All animal experiments were approved and performed according to the Guideline for the Care and Use of Laboratory Animals of Kanazawa University (permission number AP-111852).

Cell preparation
Thymus and spleen were obtained from 6-to 7-week old mice. Total thymocytes were isolated by mechanical digestion from thymus. In some experiments, thymus was digested with 0.6 mg/ ml collagenase type IV (Sigma-Aldrich) and 25 Kunitz units/ml DNase I (Sigma-Aldrich) at 37uC for 20 min. The resultant single cell suspensions were further separated by density gradient centrifugation using Histopaque-1077 reagent (Sigma-Aldrich). The cells in the interphase were obtained and were used as thymic low-density cells. Bone marrow cells were flushed out with cold RPMI 1640 medium (Sigma-Aldrich) from the femoral and tibial bones.

RT-PCR
Total RNAs were extracted from tissues using a RNeasy Mini Kit (Qiagen) and then reverse-transcribed using SuperScript III First-Strand Synthesis System (Invitrogen). PCR was done using the cDNAs, 2.

Flow cytometry (FCM)
Total thymocytes and thymic low-density cells were stained with various combinations of fluorescent dye-conjugated or nonconjugated specific Abs in magnetic activated cell separation (MACS) buffer (PBS supplemented with 2 mM EDTA and 3% FBS). For non-conjugated Abs, fluorescent-conjugated secondary Abs were used. Intranuclear Foxp3 was stained with the help of FITC anti-mouse Foxp3 staining set (eBioscience). Expression of each molecule was analyzed using FACSCalibur or FACSCanto II (BD Biosciences) with CellQuest Pro software (BD Biosciences) and FlowJo software (TreeStar).

Immunofluorescence analysis
Six or 15 mm-thick cryostat sections were fixed with cold acetone for 3 min or with 4% paraformaldehyde for 15 min and permeabilized with 0.1% Triton X for 15 min, and incubated with Protein Block Reagent (DAKO) to block non-specific binding. Then, fluorescent immunostaining was done by the standard method. After being washed with 0.05% Tween 20-PBS, slides were mounted in fluorescent mounting medium (VECTOR) with or without DAPI (DAKO). Immunofluorescence was detected in the setting that excluded the non-specific signal of the isotype control, using a fluorescence microscope, BX50 (Olympus), or a confocal laser scanning microscope, LSM510 (Carl Zeiss). DP Controller software (Olympus) and Zen 2007 software (Carl Zeiss) were used for image processing.

Uptake of antigens by thymic DCs
Alexa Fluor 488-(OVA 488 ) or 647-conjugated OVA protein (OVA 647 ) (Invitrogen) was injected into the tail vein, subcutaneous region, or peritoneal cavity of mice. FITC-conjugated dextran (M.W.; 2,000 kDa) (Sigma-Aldrich) was injected into tail vein. Thymic low-density cells were isolated at 4 hrs after OVA protein injection and analyzed by FCM. In another experiments, thymi were subjected to the immunofluorescent staining to observe the localization of antigen.

In vitro antigen uptake assay
Thymic low-density cells were incubated with 10 mg/ml OVA 647 in RPMI1640 at 37uC for 20 min. Uptake of OVA 647 was analyzed by FCM after being stained with anti-CD11c and anti-Sirpa Abs.

Generation of bone marrow chimeric (BMC) mice
Bone marrow cells were collected from WT and DO11.10 mice. DO11.10 bone marrow cells were mixed with WT bone marrow cells at the ratios of 10%, 5%, 2.5%, and 1%. WT mice were irradiated to 7 Gy by using an x-ray irradiator, RX-650 (Faxitron X-ray), and were subsequently given 10 7 mixed bone marrow cells intravenously.

Effects of OVA antigen on TCR transgenic thymocytes
OT-I, OT-II, and DO11.10 transgenic mice with or without CCR2 gene deficiency, and BMC mice were administered with OVA protein (Sigma-Aldrich) or heat-aggregated OVA protein (80uC for 10 min) in PBS through the tail vein. At the indicated time points after the treatment, thymocytes were collected and analyzed by FCM. In some experiments, the thymus was subjected to immunofluorescence analysis.

In vitro differentiation of Tregs
Thymocytes were isolated from mice 2 days after the last OVA protein injection and were incubated with 5 ng/ml rIL-7 in the presence or the absence of the indicated concentrations of rIL-2 in RPMI1640 supplemented with 10% FCS and 50 mM 2-mercaptoethanol at 37uC for 24 hrs. In some experiments, cells were preincubated with 2 mg/ml anti-CD25 or anti-CD122 blocking mAbs and were subsequently stimulated with rIL-2 in the presence of the same mAbs at 1 mg/ml.

Statistical analysis
Data were analyzed statistically using one-way ANOVA followed by the Turkey-Kramer test. Mann-Whitney's U test or Kruskal-Wallis test was used in the instances when the data were not normally distributed. p,0.05 was considered statistically significant.

Results
Thymic Sirpa + cDCs are a murine counterpart of human DC-SIGN + DCs and can efficiently uptake antigens leaking inside the IVRs In WT mouse-derived thymus, thymic Sirpa + cDCs were disseminated in the cortex and inside the IVRs, separated by two Col IV + basement membranes (Fig. 1A). Consistent with the report that DC-SIGN + DCs were present in the cortex and interlobular regions of human thymus [19,20], we detected DC-SIGN + cells in the parenchyma and interlobular region of human thymus (left panel in Fig. 1B). A double-color immunofluorescence analysis further revealed that most DC-SIGN + cells co-expressed human Sirpa molecule (Fig. 1B). Mouse thymic Sirpa + cDCs, but not Sirpa 2 cDCs, expressed the transcripts of DC-SIGN, SIGNR3, and SIGNR4 ( Fig. 1C) and DC-SIGN molecule on their cell surface (Fig. 1D), indicating that mouse thymic Sirpa + cDCs can be a counterpart of human DC-SIGN + DCs. When intravenously injected, FITC-conjugated dextran polysaccharide leaked through the vessel wall and pervaded inside the IVRs, but its leakage from small vessels was hardly observed in the parenchyma (Fig. 1E). Sirpa + cDCs in the IVRs, but not those in close proximity to the small vessels in the cortex (upper panel in Fig. 1F), efficiently captured dextran as revealed by confocal 2-D and superposed 3-D images (lower panels in Fig. 1F). Intravenously injected OVA protein was also captured mainly by CD11c+ DC population, especially by those expressing Sirpa (Fig. 1G). Moreover, thymic Sirpa + cDCs captured OVA protein, when it was administered subcutaneously, but not intraperitoneally (Fig. 1H). In contrast, CD11b + MHC class II low F4/80 + thymic macrophages failed to take in intravenously injected OVA protein (Fig. S1). Antigen uptake by Sirpa + cDCs was impaired in CCR2deficient thymus We previously reported that CCR2 2/2 mouse-derived thymus preserves the architecture of cortical TECs and mTECs with a selective decrease in the number of CD11c + CD11b + CD8a 2 Sirpa + cDCs [14]. Moreover, CCR2 gene deficiency caused aberrant localization of Sirpa + cDCs [14], especially in the IVRs where Sirpa + cDCs can efficiently capture antigens, which leaked from bloodstream ( Fig. 1 and illustrated in Fig. 2A). Furthermore, CCR2 gene ablation significantly decreased the proportion of Sirpa + cDCs capturing OVA protein among total intrathymic CD11c high DCs ( Fig. 2A). After intravenous OVA protein injection, Sirpa + cDCs of WT mice contained a substantial proportion of the cells with a higher uptake of OVA protein, but this population was markedly decreased by CCR2 gene ablation in both BALB/c and C57BL/6 mouse strains (Fig. 2B). In contrast, the cells with a lower OVA protein uptake were present with similar proportions in both wild-type and CCR2 2/2 thymus. Moreover, CCR2 2/2 Sirpa + cDCs had in vitro a similar level of capacity to capture OVA protein as WT cells did (Fig. S2). Thus, intrathymic OVA uptake by Sirpa + cDCs was impaired in CCR2 2/2 mouse-derived thymus, arising from their reduced number and inappropriate localization in the thymus.
CCR2-deficient thymus is selectively defective in the tolerogenic roles of Sirpa + cDCs with reduced intrathymic Treg differentiation and negative selection against blood-borne antigens To examine the function of Sirpa + cDCs in central tolerance against blood-borne antigens, OVA protein was intravenously injected into DO11.10 mice. OVA protein injection induced the generation of Foxp3 + Tregs in DO11.10 high mature thymocytes (Fig. 2C), which predominantly express a single TCR transgene [21]. In contrast, an increment of Foxp3 + thymocytes was markedly attenuated by the absence of CCR2 gene (Fig. 2D). It was reported that CD25 high natural Tregs were spontaneously differentiated from DO11.10 negative to low endogenous TCRrearranged thymocytes [21]. However, OVA protein injection failed to increase Foxp3 expression in this population (Fig. S3). Likewise, OVA protein injection enhanced the proportion of Foxp3 + cells in OT-II mice (Fig. S4), but not in OT-I mice of C57BL/6 background (data not shown). Moreover, IL-17 was also detected intracellularly in DO11.10 negative to low thymocytes consistent with the previous report [22], but OVA injection failed to increase DO11.10 high thymocytes expressing IL-17 (Fig. S5). When we injected the heat-aggregated OVA protein, which forms a highly immunogenic amyloid-like structure [23,24], it markedly decreased the number of thymocytes in DO11.10 mice, especially those in CD4 + CD8 + (DP) stage, and this reduction was abrogated in DO11.10/CCR2 2/2 thymus (Fig. 2E). Given selective decreases in numbers of Sirpa + cDCs and their lower capacity of antigen uptake in the thymus of CCR2 2/2 mice, thymic Sirpa + cDCs can essentially contribute to the induction of Treg differentiation and negative selection, depending on the immunogenicity of the presented antigens.

Blood-borne antigen can induce intrathymic Treg generation in collaboration with IL-2-mediated signal in a physiological condition
The repeated intravenous injection of OVA protein progressively enhanced Foxp3 expression in CD25 2 DO11.10 high thymocytes (Fig. 3A). OVA protein injection alone induced immature Foxp3 + thymocytes expressing a lower level of CD25 (Fig. 3A), but mature DO11.10 high CD25 high Foxp3 + Tregs were consistently generated by the subsequent in vitro stimulation with IL-2 after the injection with OVA protein but not PBS (Fig. 3B). The in vitro stimulation with IL-2 further efficiently increased mature Tregs following the twice injection of OVA protein, and the increase was inhibited by anti-CD25 mAb and to a lesser extent, anti-CD122 blocking mAb (Fig. 3C). Moreover, intravenous injection of OVA protein markedly increased percentage of CD25 high Foxp3 + cells among DO11.10 high cell population in BMC mice containing various percentages of DO11.10 thymocytes to polyclonal thymocytes, and the increase was correlated inversely with the chimerism ratios (Fig. 3D). Thus, monoclonal TCR transgenic thymocytes may compete with each other to receive some prerequisite signals such as IL-2 from intrathymic environment during Treg differentiation, as previously reported [25]. Thus, intrathymic presentation of blood-borne antigens which can be selectively captured by thymic Sirpa + cDCs initiates Treg differentiation in collaboration with some factors such as IL-2.
CCL2-CCR2 axis can enhance the capacity of antigen uptake by Sirpa + cDCs in tumor-bearing mice CCR2 expression by Sirpa + cDCs prompted us to examine the function of Sirpa + cDCs in tumor-bearing mice, because CCR2 ligands, particularly CCL2, are abundantly expressed in tumorbearing mice [26]. When murine colorectal cancer cell line, Col26 was subcutaneously injected into WT mice, the proportions of both DC-SIGN + Sirpa + and total Sirpa + cDCs were significantly increased among thymic CD11c high cDC population at 14 days after tumor inoculation (Fig. 4A) and were accumulated in the IVRs in tumor-bearing mice (Fig. 4B) with a marked increase in serum CCL2 (Fig. S6B). Immunofluorescence analysis revealed that CCL2 was intensively detected inside the IVRs in the thymus derived from tumor-bearing mice (Fig. 4C). The carboxy-terminal region of chemokines show a high binding affinity for proteoglycans on the vascular endothelium and in the extracellular matrix [27]. Given the abundance of proteoglycans in the IVRs in the thymus [28], serum CCL2 might be deposited inside the IVRs through binding with proteoglycan after tumor development. This assumption is supported by the observation that CCL2 was detected in the IVRs in non-tumor-bearing mice after intravenous injection of recombinant (r) CCL2 (Fig. 4C). Furthermore, tumorbearing mice exhibited enhanced intrathymic antigen uptake by Sirpa + cDCs, but not by Sirpa 2 cDCs, when OVA protein was intravenously injected (Fig. 4D). This enhanced uptake of the intravenously administered antigen was partially retarded when OVA protein was intravenously injected into Col26-7ND tumorbearing WT mice (Fig. 4E). Col26-7ND produced significantly less mouse CCL2 than the parental cells in vitro (Fig. S6A). Moreover, the in vitro co-culture of Col26-7ND reduced mouse CCL2 secretion by the parental cells (Fig. S6A). Thus, 7ND protein can inhibit endogenous CCL2 production in a paracrine and/or Sirpa molecule in region 1 (R1) and R2 are represented by unfilled and gray-filled histogram, respectively. (H) Uptake of OVA 647 at 4 hrs after intravenous, subcutaneous, or intraperitoneal injection. Percentage of Sirpa (+) OVA 647 (+) region in thymic CD11c high DCs is shown in each panel. PBS was injected as a control. Representative results from three independent experiments are shown here. doi:10.1371/journal.pone.0041154.g001 autocrine manner. Moreover, WT mice receiving the tumor expressing 7ND exhibited reduced serum mouse CCL2 concentration compared with WT mice receiving parental Col26 cells (Fig. S6B). Thus, human 7ND protein could reduce the release of mouse CCL2 by Col26 cells probably in an autocrine manner. Compared with wild-type mice, CCR2 2/2 mice exhibited depressed uptake of intravenously administered OVA protein even when they were administered with the parental tumor The numbers of Foxp3 + mature thymocytes were determined on DO11.10 and DO11.10/CCR2 2/2 mice at 2 days after OVA injection. Each symbol represents an individual mouse. Small horizontal lines indicate the mean. (E) Clonal deletion of thymocytes after OVA injection. Two mg OVA protein or heat-aggregated OVA protein were intravenously injected into DO11.10 mice. At 2 days after injection, the numbers of total thymocytes (left graph) and DP thymocytes (right graph) were determined on DO11.10 and DO11.10/CCR2 2/2 mice. Each symbol represents an individual mouse. **, p,0.01. *, p,0.05. N.S., no significant difference. doi:10.1371/journal.pone.0041154.g002 (Fig. 4E). These observations would implicate the involvement of the CCL2-CCR2 axis in the uptake of an intravenously administered antigen by Sirpa + cDCs. Intravenous injection of rCCL2 enhanced in vivo the uptake of intravenously injected OVA protein (Fig. 4E). On the contrary, rCCL2 failed in vitro to enhance antigen uptake by Sirpa + cDCs (Fig. S2). Thus, CCL2 deposited inside the IVRs can induce the trafficking of Sirpa + cDCs with a capacity to uptake an intravenously administered antigen into the IVRs, resulting in central tolerance.

Blood-borne antigen can induce intrathymic negative selection in tumor-bearing mice
To examine the tolerogenic event induced by blood-borne antigen, tumor-bearing mice were intravenously injected with OVA protein. After injection, clonal deletion in OVA-specific thymocytes, especially at DP stage, obviously occurred in DO11.10 mice bearing tumor but not in mice without tumor ( Fig. 4F and 4G). This negative selection was partially retarded in mice bearing Col26-7ND tumor (Fig. 4G). In contrast, Treg generation was depressed in the mice bearing a parental Col26 tumor (Fig. S7). Thus, tumor development can induce the shift from Treg differentiation to negative selection, together with exaggerated antigen uptake by Sirpa + cDCs.

Tumor-derived secretory antigen can efficiently induce the intrathymic negative selection
Finally, we examined whether a tumor-derived antigen can induce the intrathymic Treg differentiation or negative selection. For this purpose, we established two distinct OVA-GFP fusion protein-expressing Col26 clones; Col26-OVA and Col26-sOVA clones, which express the fusion protein only in the cytoplasm and secrete the fusion protein outside of the cells, respectively. Among the obtained Col26-sOVA clones, Col26-sOVA clone 4 secreted OVA-GFP fusion protein into the culture supernatant most efficiently (Fig. S8A) and was therefore used in the following experiments. In DO11.10 mice, Col26 cells and Col26-OVA cells grew at similar rates whereas Col26-sOVA cells grew at slower rates (Fig. S8B) as previously reported [1,2]. Accordingly, we obtained thymi at 14 days after Col26 or Col26-OVA injection and at 22 days after Col26-sOVA injection, because the tumor sizes were similar at these time points among these three groups. Compared with DO11.10 mice bearing either Col26 or Col26-OVA, both DO11.10 high mature thymocytes and DO11.10 medium immature thymocytes were decreased in the mice bearing Col26-sOVA ( Fig. 5A and 5B). This negative selection was partially retarded in DO11.10/CCR2 2/2 mice at 22 days after Col26-sOVA injection (Fig. 5A and 5B). In contrast, Treg differentiation in DO11.10 high thymocytes was marginal in mice bearing Col26-sOVA (Fig. 5C). Thus, a tumor-specific antigen can induce the intrathymic negative selection but not Treg differentiation, once it is secreted into bloodstream.

Discussion
Recently, Atibalentja et al reported the involvement of both Sirpa + and Sirpa 2 cDC in central tolerance to an intravenously administered hen egg white lysozyme (HEL) [29,30]. Because molecules with a smaller M.W. than BSA (M.W. 67 kDa) can diffuse through the blood-thymus barrier and enter the medulla, HEL (M.W. 14.4 kDa) might be captured by Sirpa 2 cDC in medulla in the previous studies [29,30]. In contrast, we observed that the Sirpa + cDCs selectively captured intravenously injected fluorescent dye-conjugated OVA protein (M.W. 45 kDa) and presented it to immature thymocytes in the thymic cortical region. Moreover, thymic Sirpa + cDCs selectively captured mouse IgG (150 kDa) [14], heat-aggregated OVA (too large to measure M.W.) [31], and dextran (M.W. 2,000 kDa), which were injected intravenously. The adult blood-thymus barrier constitutively prevents large-sized antigens from permeating into the thymic parenchyma [32]. However, we observed that dextran with a large molecular weight permeated inside the IVRs, and was subsequently captured efficiently by Sirpa + cDCs therein. Moreover, some Sirpa + cDCs capturing dextran appeared in the thymic cortical parenchyma at 18 hrs after injection (data not shown). These observations may mirror our previous observation that Sirpa + cDCs captured and conveyed OVA protein into the cortical parenchyma [14]. Thus, due to its peculiar localization in the IVR, only Sirpa + cDCs can directly capture large-sized bloodborne antigens, which are unable to permeate into thymic parenchyma.
Thymic Sirpa + cDCs captured OVA protein, when it was administered intravenously or subcutaneously, but not intraperitoneally. Boelaert J et al. previously reported that serum concentration of erythropoietin (M.W. 34 kDa) rose more slowly after subcutaneous injection than after intravenous injection [33]. They further demonstrated that intraperitoneal injection was much less effective in the raising of serum concentration than subcutaneous injection. Given the capacity of Sirpa + cDCs to capture selectively blood-borne antigens, intraperitonal injection of OVA protein may not be able to increase its serum concentration sufficient for Sirpa + cDCs to capture it in the thymus.
We demonstrated that thymic Sirpa + cDCs can capture bloodborne antigens and can initiate the antigen-specific Treg differentiation in a physiological condition. The ''two-step model'' hypothesis proposed that additional IL-2 signal is a prerequisite for intrathymic Treg differentiation after antigen presentation [34,35]. Consistent with this assumption, we observed that IL-2 induced Treg precursors to differentiate into mature CD25 high Foxp3 positive Tregs without additional TCR engagement. Moreover, the differentiation of CD25 + Foxp3 + mature Tregs was more efficient in an intraclonal low-competitive condition as revealed by the experiment using mixed BMC. Thus, Sirpa + cDCs can capture a blood-borne antigen and present it to the thymocytes, leading to . Antigen-specific Treg generation by Sirpa + cDCs in collaboration with IL-2 in a physiological condition. (A) At 2 days after the last intravenous injection of OVA protein (2 mg) into DO11.10 mice, Foxp3 and CD25 expression on DO11.10 high (R1) thymocytes was examined. PBS was injected as a control. Percentages of CD25 high Foxp3 2 , CD25 high Foxp3 + , and CD25 low Foxp3 + cells are shown in each panel. Representative results from three independent experiments are shown. (B) Expression of CD25 and Foxp3 on DO11.10 high mature thymocytes after stimulation with IL-2 in vitro. Percentages of CD25 high Foxp3 + and CD25 low Foxp3 + cells are shown in each panel. Representative results from three independent experiments are shown. (C) At 2 days after twice intravenous injection of OVA protein, thymocytes were cultured with 2 ng/ml IL-2 for 24 hrs in the presence or absence of each blocking Ab. Percentage of CD25 high Foxp3 + cells was determined. Data represent mean 6 SD from three independent experiments. *, p,0.01. (D) Percentage of CD25 high Foxp3 + cells among DO11.10 high thymocytes in BMC mice was analyzed at 2 days after intravenous injection with OVA protein. Gray-filled symbol represents OVA protein-injected mice. Data from non-BMC DO11.10 mouse are shown as a symbol of 100% chimerism of DO11.10 thymocytes. PBS (unfilled circle) and BSA (unfilled triangle) were injected as controls. Percentage of chimerism = % of DO11.10 high thymocytes in BMC thymus/% of DO11.10 high thymocytes in non-BMC DO11.10 thymus x 100. Each symbol indicates one animal. doi:10.1371/journal.pone.0041154.g003 the efficient generation of Tregs in collaboration with several factors including IL-2 in a physiological condition.
Accumulation of Sirpa + cDCs in IVRs can result in more efficient capture of an intravenously injected antigen and subsequent intrathymic negative selection. Subcutaneous tumor formation of Col26 increased the serum concentration of CCL2, which was deposited in IVRs and attracted Sirpa + cDCs therein. Consequently, Sirpa + cDCs can capture efficiently a secretory OVA protein and can induce intrathymic negative selection.
However, we cannot exclude the possibility that the secretion of a tumor-specific antigen may induce the expression of additional factors, which may induce the intrathymic negative selection in cooperation with CCR2-expressing Sirpa + cDCs.
Consistent with the previous reports [1,2], OVA-secreting tumors grew more slowly in DO11.10 mice than OVA-nonsecreting ones did, in the early phase, despite the absence of in vitro growth rate difference between OVA-secreting and OVA-nonsecreting cells (data not shown). Thus, when a tumor is was intravenously injected at 14 days after tumor inoculation. Uptake of OVA 647 in Sirpa 2 (left) and Sirpa + cDCs (right) were analyzed at 4 hrs after OVA injection. Parental Col26-, Col26-7ND-bearing mouse, and mouse without tumor were represented by a solid-lined, dash-lined, and gray-filled histogram, respectively. Representative results from 3 independent experiments are shown here. (E) Enhancement of the capability of OVA uptake by Sirpa + cDCs in tumor-bearing mice. Enhancement of OVA uptake = MFI of OVA 647 captured by Sirpa + DCs in WT or CCR2 2/2 mice bearing parental Col26 or Col26-7ND tumor/that in WT or CCR2 2/2 mice without tumor. The right unfilled column shows the fold-enhancement of OVA 647 , when OVA 647 was injected at 1 day after three times of daily injections of rCCL2 (2.5 mg/injection). (F) Two mg OVA protein was intravenously injected into DO11.10 mice at 14 days after tumor inoculation and 2 days later, expression of CD4 and CD8 in DO11.10 + thymocytes was analyzed. Percentage of DP thymocytes is shown in each panel. Data represent mean 6 SD from three independent experiments. (G) The numbers of total thymocytes (upper graph) and DP thymocytes (lower graph) are shown. Representative results from three (B and C) or four (D and F) independent experiments are shown here. Data represent mean 6 SD from four independent experiments (A, E, and G). *, p,0.01. doi:10.1371/journal.pone.0041154.g004 impalpable, a secretory OVA protein may induce intratumoral expansion of DO10.11 T lymphocytes, which can exhibit immune surveillance against an OVA-expressing tumor. In contrast, when a tumor becomes palpable, OVA protein may circulate systemically to reach thymus. Thymic Sirpa + cDCs can uptake OVA protein and subsequently induce intrathymic negative selection in DO11.10 mice. Thus, if thymic Sirpa + cDCs can be well tuned, the secretory tumor-specific antigens may become more beneficial for the immune surveillance against a tumor.
Herein, we have unraveled a pivotal role for thymic Sirpa + cDC subset in the distinct surveillance against blood-borne antigens in the central tolerance system. Moreover, we proved that an inflammatory chemokine receptor, CCR2, have profound effects on the tolerogenic capacity of thymic Sirpa + cDCs. We recently demonstrated that ablation of the CCR2 gene exacerbated the pathology of polyarthritis spontaneously developed in IL-1 receptor antagonist-deficient mice [36], suggesting its immune regulatory roles in autoimmunity. Moreover, mouse thymic Sirpa + cDCs can be a counterpart of human thymic DC-SIGN + Sirpa + DCs, the cells that are presumed to be involved in thymocyte selection in humans [19,20]. Thus, in both mice and humans, this cell population may regulate immune responses against bloodborne antigens and eventually have pathophysiological roles in the various disorders representing abnormal immune status, including cancer, autoimmune diseases, and allergy. Figure S1 Uptake of blood-borne antigen by thymic CD11b + DCs, but not macrophages. CD11b + MHC class II + DCs (R1) and F4/80 + macrophages (R3) among CD11b + MHC class II 2 cells (R2) were gated to analyze the uptake of OVA 488 (red line-histogram). Autofluorescence of each population is represented by gray-filled histogram. Representative results from two independent experiments are shown here. (TIF) Figure S2 In vitro antigen uptake by thymic Sirpa + cDCs. In vitro uptake of OVA 647 by thymic Sirpa + cDCs which were isolated from WT, CCR2 2/2 , or WT mice after three times of daily injections of rCCL2 (2.5 mg/injection) was examined. MFI of OVA 647 captured by CD11c high Sirpa + cDCs was determined and mean 6 SD from three independent experiments was shown. (TIF) Figure S3 Natural DO11.10 negative to low Tregs. At 2 days after DO11.10 mice received intravenous injection of OVA protein (2 mg), CD25 and Foxp3 expression on DO11.10 negative to low (R1) thymocytes was examined. PBS was injected as a control. The proportion of CD25 high Foxp3 + and CD25 low Foxp3 + cells is shown in each panel. Representative results from three independent experiments are shown here. (TIF) Figure S4 Induction of Treg differentiation in C57BL/6 background. At 2 days after intravenous injection of OVA protein (2 mg) into OT-II mice, expression of CD25 and Foxp3 on Va2 TCR high thymocytes was analyzed. PBS was injected as a control. Percentage of Foxp3 (+) region is shown in each panel. Representative results from three independent experiments are shown here. (TIF)